Fast High Precision Set-up of SR 620 Counter

The HP 5060 was probably the first commercial Cesium standard.  It was all analog, no microcontrollers then.  HP took over the Varian line of Cesium standards.  Then when HP and Agilent split, Agilent kept the Time and Frequency instruments and HP then became a computer and imaging company.

The FTS4060 I would call a second generation Cesium standard because it has a micro processor that replaces a lot of analog circuitry and is much easier to use.  There is a manual C field adjustment that needs to be set where the coarse thumb wheel is 1E-12 per tick and the fine wheel is maybe 1E-14 per tick.

These standards have a built in C field setting system that uses the Zeeman effect.  They turn on at a much higher level of accuracy than prior generations. The Agilent 5071A Primary Frequency Standard , Symmetricom 4065C, and others,

23 May 2006 Until about a week ago I was using the GPS 1 PPS as the start pulse and the 1 Mhz output from the FTS4060 as the stop pulse into a Stanford Research SR620 Time Interval Counter and doing 500 second averages.  There are some problems with this: The 1 MHz signal only allows a TI range of 1 micro second before rollover occurs. When you get near the rollover the average value includes data from both sides of rollover and is very wrong MOST IMPORTANT the noise is much higher than it needs to be! By changing the setup so that the FTS4060 10 MHz output feeds the SR620 rear panel Reference Input (and setting the counter to use the external reference frequency) and then using the front panel 1 kHz Reference output as the stop signal two things happen.  The rollover time is now 1 milli second (10,000 times longer) and the noise is reduced (probably SQRT(10000) = 100) by a huge amount. With a 1 MHz stop if the TI is between 0 and 200 or between 800 and 1000 there is a chance of rollover points being in the average and between 0 and 100 or between 900 and 1000 it's almost certain that there will be rollover points in the average.  Because of this I'm currently slewing s/n 1003 which was at 980 ns and may have an optimum C-field setting of 908 by setting the C-field at 000 where the slew rate may be in the +20 to +40 ps/sec area so it sill take many hours to get the TI to about 500 ns. Note that the 1 kHz out and the cable between Ref Out and B in, have an associated time delay so the TI numbers will not match those with a direct connection.

17 May 2006 s/n 1227 By plotting the offset vs. C-field switch setting it's clear that the slope is -1E-13 per tick.  This is also a great way to see which data points are not valid.  For example the old data point for C-field setting 492 was +9.8E-13 which is maybe 10 times higher than where it may end up.  As of 22 May 2006 it's -6.24E-14 with R2 of 0.84.  When R2 gets up to one or two nines, then we'll see.

Note that a Cs frequency source is just like any other high stability source and needs to have it's frequency set.  The big advantage of Cs is that once set it will not drift like Rb or Quartz.  Note this is because of the defination of time, and maybe is not reality. 

15 Feb 2005 - The time interval plot was drifting up so an adjustment was made to the C field.  It turns out that the adjustment was too large.  24 hours after the adjustment for abut 6 hours the time interval stayed constant within about 10 nano seconds.  This indicates that even 24 hours after a C field change the frequency has not stabilized.   It really does take two to three days for the unit to stablize after a C-field change.

Slew using C Field

9 March 2006 - When using a time interval counter where the start signal is the 1 PPS output from a GPS timing receiver and the stop signal is from the 1 MHz output of the FRS4060 the counter has a range of 0 to 1,000 nanoseconds.  If the time interval rolls over you get a saw tooth type plot.  In order to slew the time interval away from 0 or 1,000 you can set the C field as far as possible from the correct setting.  For example on s/n1013 the correct setting will be near 855, so setting to 000 causes the frequency to slew at about 1,400 ns per day which is  an offset of 1.6E-11 but you can see to get a 500 ns change will take about 8.5 hours.  Not as fast as you would like.  There may be other ways to slew, but so far this works.

Note that when the C field is misadjusted as much as possible (1.6E-11) the offset is 10 to 100 times better than the best ovenized lab grade crystal oscillators daily aging rate(1E-9 to 1E-10).

Thumb Wheel Switch Sensitivity

11 Dec 2005 - s/n 1013 after complete disassembly and reassembly seems to be working, not like in Jan 2005.
With C field thumb wheel switches set for 500 the plot of (start = GPS 1 PPS, stop= FTS4060 1 MHz zero crossing) has a positive slope of about 5.7864 ps/sec and the plot with C field at 900 has a slope of -1.7764 ps/sec.  So the thumb-wheel switches may have a scale factor of:
Scale Factor = (7.5628 - (-1.7764)) / (900 - 500) = 0.023348 ps/sec or about 2E-14 per click.
See framed coment above the scale factor for s/n 1227 is -1E-13 per tick.

Test Signal

GPS

One way to adjust the C field 3 digit thumb wheel pot is to use GPS.  Although there is some jitter on the GPS 1 PPS signal that amounts to maybe plus and minus 50 ns (Motorola 8 chan), the accuracy over a 24 hour period is on the order of 1E-12.  The Motorola M12+T has about 9 ns and so is much better.  The GPS receiver should be used directly.  Also there a lot of jitter on the 1 MHz FTS4060 output, much better to divide it down to 1 kHz or 1 PPS.

It seems that the scale factor for the /S24 using is 1E-13 per count NOT the 2E-14 in the normal FTS4060 manual.
  9 Jan 2005 - Plot -
10 Jan 2005 - Plot of nano seconds of Time Interval vs seconds of running time for s/n 1013 - 5.9E-12?
12 Jan 2005 - Plot - Between 124,920 and 169,680 (12.4 hours) the 1 PPS input to the PRS10 was removed and when reconnected caused a negative swing that lasted until 248,400 seconds.  But it appears that a C-Field setting of 913 is pretty close to correct.  The drift is in the e-13 ot E-14 area.
18 Jan 2005 - Plot of s/n1013 vs. s/n 1227, 1227 vs GPS & 1013 vs GPS, now using s/n 1227 as 10 MHz ref for SR 620 counter.
s/n 1227 has it's C Field set at 600 as received from Govt Liq. and it appears to be moving at -3.6E-12. s/n 1013 seems to be moving at 5E-10 more like an OCXO than a Cesium, but the Lock LED is on and the beam current peaks as it should.  What wrong?

1 Feb 2005 - s/n 1227 - I tried to use the time interval between GPS and the 1 Mhz output to set the C Field by getting the offset and then dialing in the correction (it looked like the three thumb wheels were 1E-12, 1E-13 and 1E-14), but the resulting slope after a few days of observation seemed to overshoot.  A better way would be to use a binary search where at each attempt you would half the error.  I think I have the setting to within a single count on the finest wheel, but it'll take some days to see.

10 Feb 2005 - s.n 1227 - Still have C field at 544.  The 10 day plot shows GPS wandering within a 150 ns range so the poorest stability might be 1.7E-13, but an average would be more like parts in E-14.

10 Feb 2005 - Enabled Ionospheric correction in GPS receiver and the delta time jumped up to the 500 ns range, so this may account for the 100 or or ns variation the last 10 days.  More time will tell.
11 Feb 2005 - changed GPS to track 4 highest satellites and changed elevation mask to 30 degrees.
28 Feb 2005 - The C field has been at 570 for about 9 days and on average there does not appear to be any drift, but it's difficult to tell.

2 March 2005 - To improve the stability of the GPS 1 PPS I increased the elevation mask again, this time from 30 degrees to 50 degrees.  It has made a big improvement.  The standard deviation after 1,000 seconds worth of 1 PPS averaging is now in the 30 ns area where before it was in the 200 ns area.  During 3 days of observation there never was a time when there were no satellites above 50 degrees.  Since I'm running the GPS receiver in the timing mode (known antenna position) only one satellite is needed for a timing solution.

8 March 2005 - C filed at 565 - After the problem with the 4060 going crazy after a beam current centering.  Needed to cycle power to get good operation.

13 March 2005 - Yesterday the counter got unplugged, but neigher the FTS4060 nor the Austron 2100T were unplugged.  Both of these instruments have warning LEDs that would indicate a loss of mains power, but the FTS-4060 output frequency became more unstable after this event.  This morning I unplugged the FTS4060 for 10 seconds and restarted it.  After that the standard deviation on the time interval improved from over 300 ns to more like 30 ns.  Maybe there are some power supply caps that need replacing or more caps need to be added?

Also the amount of averaging on the GPS 1 PPS needs to be increased.  At 1,000 averages the best stability that can be seen in one day is about
(3 * 35 ns * 2)  / (SQRT(1,000) * 86400) = 7.6E-14, but by going to 5,000 seconds the system improves to 3.4E-14.  So starting a new plot.

15 April 2005 - Switched to an SynPaQ/III with Motorola M12+T GPS receiver.  This unit has 3 to 4 times less variation than the old 8 channel UT+ GPS receiver.  But there appears to be a parabolic change in the plot over the past 5 weeks that I don't understand.  The C filed has been at 568 since 20 March 2005.

28 April 2005 - the plot for s/n 1227 vs both GPS and Loran-C still appears to be parabolic, indicating some type of aging which is NOT supposed to occur whith a Cesium source. Aging is about -3E-14 per day.

29 April 2005 - the aging rate seems to be slowing down.  It's now -2.2E-14/day.

1 Feb 2006 - s/n 1013 seems to be working after having all the modules taken apart (working on technical manual) and then put back toghther again on 9 Dec 2005.  Changing the C-field causes a change that takes about a week to settle down (now C=850) and for the last few days the 1 PPS has stayed within about 1 ns of the Motorola M12+T pulse (maybe 1ns/3 days = 4E-14).

6 Feb 2006 - s/n 1013 is showing drift like s/n1227.  The equation for s/n 1013 is:
 y = 2.7943x² - 302.64x + 8969.4 and the quality of fit is
R2 = 0.9088.  The x-axis is in days and the y-axis is in ns. 
The first deritive of the equation has a first term of 2 * 2.7943 * x ns/day or +5.3E-14 drift rate.

I don't know if this is a measurement problem or a problem with the FTS4060 standards.

9 March 2006 - The apparent parabolic aging was a measurement problem related to setting the time interval counter trigger level improperly (50 Ohm source and load TTL should be at 1.25 Volts, NOT 2.4 volts).
Now s/n 1013 is looking very good.  Another problem may have been that the Ultra Stable Oscillator coarse frequency was not set properly.  It now has been centered and now looks like +4E-13 offset which I'm trying to adjust to be much better.

28 April 2006 -

Note: GPS has some noise.  For example the Motorola M12T+ has a standard deviation of about 9 ns when 500 Time Inteval readings are averaged (reference is some good oscillator).  So you might expect that the noise will peak +3 sigma and -3 sigma from the mean value.  This means that the offset you can see is about 54ns/(measurement time in seconds).

Meas time range	observable offset
1 to 4 min	1E-9
15 min to 1 hr	6E-11
1  to 4 hr	1E-11
12 to 48hr	1E-12
1 - 4 days	6E-13
week to month	1E-13

When plotting Time Intervals in Excel you can fit a trendline and also get an R squared quality of fit number.  R^2 should be some number of nines for a good fit.  If it's not then there's something wrong.

17 May 2006 - there are times that last for about a couple of hours whee the SR620 is displaying a standard deviation for a 500 second average as high as a few hundred nano seconds.  I still don't know what causes this.  Some possible things that might cause it are:

• multipath may cause a problem if there was only one or two sats visable, but I would think with 3 or more visable a poor satellite would not cause a problem?
• no satellites at all would allow the GPS receivers 1 PPS to be coming from it's raw crystal
• some problem with the SR620 - I have disconnected the PRS10 as the external ref since it's not needed and between the PRS10, it's power supply, GPS receiver there's just that much more to go wrong.  But this has not seemed t make any difference.

Rollover The problem was that the data was getting near the 1,000 uS rollover point caused by using a 1 Mhz signal for the counter B (stop) input.  SR620 Trick Instead of connecting the cesium 1 MHz signal to the B input, connect the 10 MHz signal to the counter's rear panel 10 Mhz input.  Use SEL, SET & SCALE^V to enable rear clock input.  Then connect the  front panel 1 kHz Reference TTL output to the B (stop) input.  Now the rollover will be every 1 mS, or a thousand times improvement.

LORAN-C

Zeeman Resonance

The HP 5060A manual says the Zeeman frequency should be 42.82 kHz and about 1 volt RMS.  And that an error in the Zeeman signal of 1% translates into a Cs frequency error of 3.6E-12, so it needs to be set to within about 1 Hz.  The amplitude of the Zeeman signal and the C-Field can be adjusted, with the modulation off and the loop open to set the C-Field, or the C-Field can be measured by adjusting the frequency and amplitude of the Zeeman input to get maximum beam current.

Mr. Pieter Zeeman won the Nobel prize in Lecture 1902 along with Mr. Lorentz for explaining a splitting in the spectral lines of light caused by magnetic fields.  This explanation was based on the new things called "electrons", but did not take into account quantum effects like up and down spins.  His experiments and the theory by Lorentz shed a lot of light on what an "electron" was.

So far I don't have an audio generator that has the required frequency settability AND enough drive power to do this test.

Corby D Dawson and Tom Van Baak have described how the audio frequency for the Zeeman effect depends on the physics package and I'm rephrasing it here.  The definition of the second is based on a Cesium standard running in a zero magnetic field at sea level with a frequency of
9192.631770 MHz.  Real Cesium tubes run with a very small magnetic field and so their frequency is slightly off that for a standard second, but the manufacturer knows how far off and allows for it so that the final 10 MHz or 1 PPS is exactly correct.

Tube Model	C field milli G 1	Synth 2 Freq Hz	Tube Freq Hz	Zeeman Hz
5060A
5061A	61	12,631,771.6	9192 631 771.6	42.82
5061B	76	12,631,772.5	9192 631 772.5	53.53
5062C	100	12,631,774.3	9192 631 774.3	70.40
FTS4060

Note ¹ - The C field coils in both HP and FTS Cesium tubes have the same milli gauss per milli amp constant and so the C field is determined by how the main frame is setup. 
Note ²- The synthesizer that generates the frequency that's fed to the multiplier is also in the main frame and has a frequency that matches the strength of the C field.

Note that as long as the C field and synthesizer are matched to each other the system should work properly.
There may be an error reversing the synthesizer frequencies of the HP5061A and HP5061B

Dead New Batteries

At first one of the new batteries not only would not put out any voltage, but actually had reverse voltage across it and the 4060 was still running.  This means that the switching supply will keep the 4060 going on less than 12 volts (although it may or may not start a cold 4060 on that low a voltage).  The "bad" battery looked just like the good batteries when connected to the charging power supply. 

So now I have made up a simple 12 Volt battery checker that's just a number 1156 car tail light bulb soldered to a clip lead that was cut in half.  This pulls a couple of amps to light brightly and with only 1 amp will take some seconds to light dimly.  This works much better than the Radio Shack 22-080 battery tester that shows a dead battery as good.

Note that the very common 12 V 7 AH batteries come with both 1/4" (0.250") and 3/16" (0.187) quick connect type terminals.  On the batteries I got some are 1/4" and some are 3/16".  So you need to check each battery, even though at a quick glance they look the same.

"Moreover, in rubidium frequency standards, the available C-field adjustment range limits the useful life of the unit. For example, in one of the most popular rubidium frequency standards currently on the market, the manufacturer provides a C-field adjustment range equivalent to +1.5.times.10.sup.-9. The aging rate of the standard is specified as 2.times.10.sup.-10 per year. Consequently, at the specified aging rate, the limited C-field adjustment range limits the useful life of this rubidium frequency standard to 1.5.times.10.sup.-9 /2.times.10.sup.-10 =7.5 years."

Note It is an easy job to replace the 5030 Assembly and that may allow using the complete 5030 Assembly from a working /S24 unit to bring a defunct FTS4060 back on line.  This can be done in a few minutes.  But I don't know where the additional modules are located on a full featured 4060.  If they are on the right side ( the 5030 is on the left side) then it would be very easy.  If they are in the way of removing the 5030 Assembly then it would take longer.
On the upper left is the 10 MHz OCXO. 

A1A5 Dist AMP

At the upper right is the A5 Distribution Amplifier.  This amy be an A5/S24 with the front 10 Mhz output missing. 

A1A5 & A1A7 Sub Assembly

A1A2 Mother Board

A1A6 Ultra-stable Oscillator

A1A7 x18 Mult & Mixer 

Just under the A5 Amplifier is the Times 18 Frequency Multiplier (10 MHz in, 180 MHz out) and mixer.  

As seen in the photo the connectors are: 10 MHz in,  connector with Black, Gnd, and Red wires going to J4 on the A3 Microprocessor board.  Cable with Black, Red, Green (ground) & blue wires soldered to feedthroughs going to connector J4 on Cesium Beam Tube motherboard.. 12.6 MHz input & 180 MHz output.

A1A8 Cs Power Supply

In the lower right corner is the A8 Power Supply for the Cesium Beam Tube that includes the two HIGH VOLTAGE outputs.  The bottom of this PCB is visible at the top left front when the top cover is removed.  You might not want to have your hands anywhere near this board when power is applied.

A1A9 DC Input Filter

Although this PCB is in a slot and should be able to slide out, it's trapped by the female thread fitting used to attach the 5030 sub assembly to the chassis.  It has a dual electrolytic cap, a single electrolytic cap, a diode and an inductor.

A2 RF Assebmly

At the bottom center is the 3 x 7" A2/S24 RF Assembly (56219-05280-011 Assy 05281) that takes in 10 MHz and outputs 1 MHz.  On a full featured 4060 this board would also output 100 kHz and 10 MHz. 

It may that  the 74LS90 divide by 10 circuit could be bypassed to allow two 10 MHz outputs instead of the two 1 MHz outputs that are on the /S24 versions.

A1A3 Micro Processor Board

A1A4 12.6 MHz Synthesizer

Directly under the A2 board is another same size board, the A1A4 12.6 MHz Synthesizer (56219 Assy) that shares the same 14 conductor ribbon cable and has a single coax cable that goes to the times 18 Multiplier.   There are a half dozen Synchronous Four-Bit Counter 54LS161 ICs on this board.

A1A1 Cesium Beam Tube Assembly (Physics Package)

Interface PCB

PCB Rails but no Card Edge Connectors

The PCBs are held by rails, but there are no card edge connectors on the PCBs.  All connections are made by coldered wires, coax connectors (typcially standard 50 Ohm SMB), or rectangular connectors.  There is an unused pair of rails above the microprocessor PCB, but if another board is used there it would ned to have notches to clear the to coax connectors coming from the uP board.

26 Feb 2006 - Note after almost getting the C field set on s/n 1013, there was a power failure lasting about 2 seconds.  But 1013 had been running for many months and was working well.  After the power came back on the lock LED did not light.  After giving it about 4 hours still with no lock light, the above procedure was used to set the control voltage and loop gain, and afterwards the lock LED turned on as soon as the closed loop and mod on switches were thrown.

28 Apr 2006 - You really can only center the Control Voltage using the front panel controls (as described above in the OPEN mode).  When there's a continuous search or the yellow monitor LED is on with the green LOCK LED, then it's time to adjust the coarse frequency of the crystal oscillator.  This can be done with the source locked.  Remove the bottom cover and with the monitor switch in position 4 (CONTROL) note the reading.  In my case it was over 3.  Then adjust the coarse pot on the USO to bring the needle a little to the other side of 2.5, in my case to 1.6.  When the control voltage gets closer to 2.5 the yellow monitor LED will go out leaving only the two green LED for LOCK and AC power.

Hint: leave the door slightly open by turning the screw all the way out, closing the door and then turning the screw 1/4 turn.  This way if the ALIGN lamp turns on you can see it.  You could leave the door open, but for me it's in the way of other stuff.

The manual implies that you can make the loop gain or beam current adjustments on the fly.  But on the two occasions that I have tried to do that the FTS4060/S24 gets confused.  The fix has been to pull the line plug for about 15 seconds and restart.  Without the restart the control loop is oscillating.

Recommendation